Open Access

Nils Floß

• Key performance indicators form the basis for controlling large production systems and enable decision-makers to derive appropriate measures. Currently, large production companies are implementing extensive digitization measures for the collection and visualization of these key performance indicators. However, the potential of predictive analytics and automated decision-making has only been exploited in isolated cases. The potential for improving the quality and efficiency of decision-making processes is not being exploited. This means that an opportunity to reduce production costs remains untapped. In addition, existing approaches are limited to individual solutions that are poorly scalable, adaptable, and reusable. The problem is solved by developing a methodology using the design science research approach. In a requirements analysis, the findings from the literature review are compared with the practical requirements of large production systems in order to create a detailed catalog of requirements. Based on this, the sub-steps of the methodology are developed and orchestrated into an overall methodology. Validation is carried out through the implementation of concrete use cases, comparison with predefined success criteria, and incorporating expert feedback. The methodology is used to develop forecasting capabilities and decision automation and integrate them into the daily routine of production control in order to optimize the cost efficiency of production. The scalability and reusability of the applications developed with it enable their cross-plant introduction and uniform adaptability. As the first use case, an automated cloud-based end-to-end data pipeline for container prediction is being developed in the supply centers of the BMW plant in Munich, enabling dynamic personnel planning and cost optimization.